One degree micro-macro manipulator integration test

aIR

A 111 Ob M3D671

NIST

PUBLICATIONS

One Degree Micro-Macro
Manipulator Integration Test

Richard J. Norcross

U. S. DEPARTMENT OF COMMERCE
Technology Administration
Intelligent Systems Division
National Institute of Standards
and Technology
Gaithersburg, MD 20899

■ IWWi 'il '"TCg..

NIST CENTENNIALS

NIST

National Institute off Standards
and Technology

Technology Administration
U.S. Department of Commerce

NISTIR 6562

One Degree Micro-Macro
Manipulator Integration Test

Richard J. Norcross

U. s. DEPARTMENT OF COMMERCE
Technology Administration
Intelligent Systems Division
National Institute of Standards
and Technology
Gaithersburg, MD 20899

August 2000

U.S. DEPARTMENT OF COMMERCE
Norman Y. Mineta, Secretary

TECHNOLOGY ADMINISTRATION

Dr. Cheryl L. Shavers, Under Secretary

of Commerce for Technology

NATIONAL INSTITUTE OF STANDARDS

AND TECHNOLOGY

Raymond G. Kammer, Director

ONE DEGREE MICRO-MACRO MANIPULATOR INTEGRATION TEST

For the

Automated Paint Application, Containment and Treatment System

(APACTS)

for:

Naval Surface Warfare Center
Carderock Division
9500 MacArthur Blvd.
West Bethesda, MD 20817-5700

Richard J. Norcross

Intelligent Systems Division

National Institute of Standards and Technology

Gaithersburg, MD 20899-8230

Disclaimer

No approval or endorsement of any commercial product by the National Institute of
Standards and Technology is intended or implied. Certain commercial equipment,
instruments, or materials are identified in this report to facilitate understanding. Such
identification does not imply recommendation or endorsement by the National Institute
of Standards and Technology, nor does it imply that the materials or equipment identified
are necessarily the best available for the purpose.

Copyright

This publication was prepared by United States Government employees as part of their
official duties and is, therefore, a work of the U.S. Government and not subject to
copyright.

Acknowledgement

This report is partial fulfillment of sub-contract AM 02-9802001 with AmDyne
Corporation of Millersville, Maryland.

One Degree Micro-Macro Manipulator Integration Test

ii

August 21, 2000

Abstract

The Carderock Division of the Naval Surface Warfare Center is developing the Automated
Paint Application, Containment, and Treatment System (APACTS) to apply anti-
corrosive and anti-fouling paints in an environmentally sound manner. To provide
accurate motion over a very large volume, the APACTS motion system employs a self-
propelled base carrying a long reach macro-manipulator which in turn carries a quick
response micro-manipulator to maneuver the paint nozzle and containment device along
the ship hull. The manipulators run separate but coordinated trajectories whose
combination is the path of the paint nozzle. Based on sensors and feedback from the
operator’s observations, the micro-manipulator’s trajectory is shifted to keep the paint
nozzle at the appropriate position relative to the surface being painted. The micro-
manipulator communicates the shift to the macro-manipulator, which adjusts its
trajectory to remove the error and re-center the micro-manipulator. This report
investigates the minimum requirements for the interface between the macro and micro-
manipulators. The investigation includes experiments that test the micro-macro interface
by observing the system's response to induced errors. The results indicate an interface
reporting the size of the shift and the frequency of the report is sufficient to control the
manipulator system, but may not be sufficient for the APACTS application. The report
includes possible improvements to the interface.

One Degree Micro-Macro Manipulator Integration Test

August 21, 2000

DRAFT

Table of Contents

1 Summary 5

2 Introduction 6

3 Methods, Assumptions, and Procedures 8

3.1 Manipulators 9

3.2 Trajectory 10

3.3 Operator Interface 1 2

3.4 Micro-Macro Interface ! 3

4 Results and Discussions 14

5 Conclusions 1 8

6 Recommendations 1 8

6. 1 Continuous Feedback 1 8

6.2 Surface Modeling 19

7 References 1 9

List of Figures

Figure 1. APACTS System Concept 5

Figure 2. RCS Hierarchy 8

Figure 3. Modified ATR-60 AWP as Macro-Manipulator 9

Figure 4. Micro-Manipulator on AWP Basket 10

Figure 5. Micro Manipulator Trajectory 1 1

Figure 6. Trajectory Speeds to Simulate Curved Surface Adjustments 1 2

Figure 7. Monitor Screen 1 3

Figure 8. Sample Test Run 1 5

Figure 9. Correction with Initial Speed Error 16

Figure 10. Correction with Initial Position Error 16

Figure 1 1. Simulated Surface Following 17

Figure 12. Relocation Sequence with Excess Boom Speed 18

One Degree Micro-Macro Manipulator Integration Test August 2 1 , 2000

IV

1 Summary

The Carderock Division of the Naval Surface Warfare Center is developing the
Automated Paint Application, Containment, and Treatment System ( APACTS) to apply
anti-corrosive and anti-fouling paints onto Navy ship hulls in an environmentally sound
manner. APACTS’ motion system (Figure 1 ) employs three motion components; a self-
propelled, repositionable base, a long reach macro-manipulator, and a quick response
micro-manipulator to move the paint nozzle and containment device along the hull
surface. The combined motion trajectories of the macro and micro-manipulators form the
paint application trajectory. This report investigates the interaction between these two
manipulators.

Figure 1. APACTS System Concept

One Degree Micro-Macro Manipulator Integration Test

5

August 21, 2000

The micro-manipulator trajectory is a repetitive sequence of positions that traverses only
a portion of the micro-manipulator’s range in any direction. That portion of the micro-
manipulator's volume that is not used by the trajectory is called the excess volume. The
operator removes nozzle position errors by shifting the micro-manipulator’s trajectory
into the excess volume. For example, if the micro-manipulator can reach between 10 and
30 cm in a given direction, and the trajectory requires only 10 cm in that direction, then a
nozzle position error of 3 cm can be removed by having the micro-manipulator’s
trajectory operate between 1 8 and 28 cm rather then between 1 5 and 25 cm. The micro-
manipulator periodically sends the accumulated shifts (i.e., the offset) to the macro-
manipulator. The offset triggers a change in the macro-manipulator trajectory that
reverses the position errors and results in the micro-manipulator being re-centered.

To test the system’s ability to coordinate macro and micro motions in response to
corrective inputs, experiments were conducted where two types of known errors were
introduced into the macro manipulator’s trajectory. In the first set of tests, the macro-
manipulator motion begins a small distance from the intended start position to produce a
tool position error. In separate tests, changes to the macro-manipulator’s velocity, either
in the beginning or middle of the run, produce a tool speed error. The experiments verify
the interfaces are sufficient to coordinate and stabilize the two manipulators. However,
the minimum micro-macro interface suffers periodic position errors that may cause gaps
in the paint coverage and be unacceptable for the APACTS application.

A review of the test results indicate the overall system performance may be improved
through enhancements to the detection and communication interfaces. In these
experiments, an operator detects the position error. The operator was unable to observe
the proper nozzle position through much of the micro-manipulator’s trajectory cycle.
Speed errors during this time accumulate into significant position errors. Improvements in
position detection, either through improved operator assistance or through an automatic
sensor, may avoid error accumulation and improve overall system performance. The
micro-macro interface used in these experiments does not provide the macro-manipulator
with the curvature of the surface. Since surface curvature affects the required macro-
manipulator speed, curvature data may allow the macro-manipulator to more accurately
and rapidly adjust its velocity in response to curvature changes. The value of these
enhancements must be verified by separate tests.

2 Introduction

To guard against the harshness of the sea, ships are covered with anti-corrosive and anti-
fouling paints that must be periodically replaced to maintain their effectiveness. During
replacement, hazardous airborne particles (HAPs) are inadvertently discharged into the
environment diminishing air quality, and endangering shipyard personnel and the

One Degree Micro-Macro Manipulator Integration Test

6

August 21 , 2000

surrounding harbor. The Carderock Division, Naval Surface Warfare Center (CD-NSWC),
Environmental Quality Department conducts research and development leading to fleet
implementation of pollution-control materials, processes, and equipment that enable
Navy ships to be environmentally responsible. CD-NSWC is responsible for providing
the Navy with the technical expertise to solve existing and emerging waste management
problems. Pursuant to that responsibility, CD-NSWC is developing the Automated Paint
Application, Containment, and Treatment System (APACTS) to significantly reduce
HAP discharge from the painting operation ( 1],

The primary components of APACTS are the delivery system, the containment system,
the treatment system, and the manipulation system. The delivery system consists of a
paint mixer, strainer, sprayer, nozzle, and associated equipment. The containment system
surrounds the paint sprayer and includes a capture shroud, recovery vacuum, hoses, and
controls. The treatment system includes waste transport, waste isolation, filter elements,
and their support equipment. The manipulation system consists of those devices that
move the sprayer and containment shroud. The components complement each other to
produce an effective, economic, and environmentally-sound system.

The Intelligent Systems Division of the National Institute of Standards and Technology
(NIST-ISD) supports APACTS development through investigation of new and existing
technologies to carry, maneuver, and manipulate the APACTS sprayer and containment
system. Since a single manipulator would be unable to achieve the performance
requirements at an acceptable cost, APACTS uses a series of three manipulators to
position the system about the dry-dock, to reach along the hull, and to maintain proper
standoff and motion. After the mobile base positions APACTS in or around the dry-
dock, a long reach, but slow response macro-manipulator carries a high accuracy, fast
response micro-manipulator to simultaneously provide sufficient reach and accuracy.

The combination of dissimilar manipulators is known by several names including; macro-
micro, macro/micro, maxi-mini, and major-minor. Many researchers have investigated
macro-micro control ( [2]-[8] ). These approaches rely primarily on either a well-defined
trajectory or a well modeled pair of manipulators. While none adequately address
problems of working throughout a very large volume in a poorly defined environment,
several micro-macro control strategies may be extensible to the APACTS problem.

The proposed APACTS controller follows the hierarchical control theory of the Real-
time Control system (RCS) [9], Under RCS, tasks are spatially decomposed along the
branches of the hierarchy and temporally decomposed across the levels of the hierarchy.
Each controller node contains a Sensor Processing module (SP) that interprets sensor data,
a World Model module (WM) that maintains the model of the control system's
environment used to plan task sequences, and a Behavior Generation module (BG) that
plans and executes the task sequence (Figure 2). An RCS controller treats the

One Degree Micro-Macro Manipulator Integration Test

7

August 21 , 2000

manipulators as independent systems with a common supervisor that coordinates the
subordinates through the initiation of related commands. Cooperation within a task
occurs via the hierarchy’s World Model. Thus, to the macro-manipulator, the micro-
manipulator is effectively a pre-processed sensor. The micro-manipulator’s position
within its work volume is an offset value that the macro-manipulator subsequently
minimizes with changes to its own motion. As the offset value diminishes, the micro-
manipulator returns to the center of its volume.

Figure 2. RCS Hierarchy

This report reviews the interaction between the macro and micro manipulators. The work
here intends to establish sufficient requirements for an interface between two serially
linked and cooperating manipulators. We demonstrate that two manipulators, with large
differences in their servo frequencies, can be coordinated through an interface operating at
a lower frequency then the slower manipulator.

3 Methods, Assumptions, and Procedures

The experiment tests the coordination of two serially linked manipulators. The two
manipulators execute separate trajectories that, when simultaneously executed, step and
maintain a tool point over periodic marks on a vertical surface. We introduce position and
velocity errors to the macro-manipulator trajectory to evaluate the system’s ability to
correct the errors through the macro-micro interface. By way of a camera on the micro-
manipulator, an operator observes the effect of the error on the tool position and, via a
joystick, offsets the micro-manipulator’s trajectory to compensate. The micro-

One Degree Micro-Macro Manipulator Integration Test

8

August 21, 2000

manipulator communicates the offsets through the micro-macro interface and the macro-
manipulator adjusts its trajectory to re-center the micro-manipulator.

3.1 Manipulators

The macro-manipulator is an ATR-60 aerial work platform (AWP) from Snorkel, Inc. of
St. Joseph, MO. The standard AWP has digital proportional valves on several of its
actuators. The experiment’s AWP has similar digital proportional valves on all actuators
and each actuator is fitted with absolute position and relative motion sensors. Servo
control modules monitor the actuator motion and adjust the oil flow through the valves to
cause the actuator to follow a motion path. A supervisory controller coordinates the
actions of the servo modules such that the AWP's basket can follow Cartesian paths or a
surface as shown in Figure 3. The macro-manipulator's supervisory controller updates the
goal position at 8 Hz. The actuator controllers close the actuator servo loop at 30 Hz.

Figure 3. Modified ATR-60 AWP as Macro-Manipulator

The micro-manipulator in the experiment is a one axis linear actuator carrying a small
video camera (Figure 4). The camera provides position feedback to the operator during the
experiment. The manipulator is controlled by a Smart Motor from Anamatics, Inc. of
Carlsbad, CA. The Smart Motor runs in Anamatics' extended cam mode and contains a
repeating trajectory. The controller shifts (or offsets) the trajectory in response to signals
on an analog input port. The micro-manipulator controller updates the goal position at
200 Hz and closes the motor's servo loop at 4 kHz.

One Degree Micro-Macro Manipulator Integration Test

9

August 21, 2000

Figure 4. Micro-Manipulator on AWP Basket

3.2 Trajectory

The experiment uses the vertical component of the vertical compensation trajectory
discussed in [10], Under the vertical compensation trajectory, APACTS paints in vertical
swaths, where a swath is a set of horizontal stripes painted sequentially from top to
bottom. The test trajectory imitates an APACTS application with a 30 cm (12 in)
effective spray width (aligned vertically), a 36 cm/s ( 14 in/s) nozzle speed (applied
horizontally), and 240 nr/h (2600 ft2/h) production rate. The nominal macro-manipulator
trajectory moves the basket vertically down a wall at 5 cm/s (2 m/s). In order to maintain
the vertical position during paint application, the nominal micro-manipulator trajectory
moves up 19 cm (7.5 in) for 3.75 seconds (5 cm/s (2 in/s)) (the upstroke) then returns to
the initial position in the subsequent 2.25 seconds (the relocation stroke). The first loop
begins with a 3 cm/s2 acceleration (Figure 5).

Known errors injected into the macro-manipulator trajectory test the system's stability.
The error conditions include an initial position error and an initial velocity error. A
simulated surface with abrupt curvature changes tests the system through a sequence of
sudden velocity errors.

One Degree Micro-Macro Manipulator Integration Test

10

August 21, 2000

Figure 5. Micro Manipulator Trajectory

Boom speed modifications simulate changes to the surface curvature. Along a flat surface,
the micro-manipulator, whose reference is on the surface, and the macro-manipulator,
whose reference is some distance off the surface, move at the same speed. However along
convex surfaces, the macro-manipulator must move faster to maintain the proper relative
position. Thus abrupt boom speed changes simulate an abrupt shift from a flat surface to
a curved surface. With a 1 83 cm (72 inch) surface radius and a 91 cm (36 in) standoff, the
speed change is 50% of the nominal speed. To follow an actual wall transiting from a flat
to a convex surface, the boom speed would increase from 5 cm/s to 7.5 cm/s (2 in/s to 3
in/s). To simulate this action, the boom speed is reduced 2.5 cm/s ( 1 in/s). Similarly, a
boom speed increase from 5 cm/s to 7.5 cm/s (2 in/s to 3 in/s) simulates a transition from
a flat surface to a concave surface. Similar boom speed changes simulate other transitions.

The simulated test surface (Figure 6) is flat for 61 cm (24 inches), convex (at 183 cm
radius) for 144 cm (56.5 in), flat for 30 cm ( 12 inch), concave (at 183 cm radius) for 144
cm, and finishes as a flat surface. The transitions generate abrupt 2.5 cm/s changes in the
boom speed which create position errors that are removed by the operator’s input. The
anticipated changes for an APACTS application are less severe then those used in the
experiment.

One Degree Micro-Macro Manipulator Integration Test

August 21 , 2000

Figure 6. Trajectory Speeds to Simulate Curved Surface Adjustments

Surface Following
Speeds

Flat Surface
Adjustments

-2.5 cm/s

+2.5 cm/s
+2.5 cm/s

-2.5 cm/s

f

3.3 Operator Interface

The macro-micro control concept relies on observations of the task to determine part of
the motions of the micro-manipulator. Since the position based on actuator sensors and
kinematics calculation is unreliable, other sensors must detect the relative pose of the tool
point to the task. While the full APACTS application may use automatic sensors to
generate the information, an operator generates the appropriate feedback signals for this
experiment.

The experiment’s operator interface consists of the video signal from the micro-
manipulator's camera and a one degree of freedom joystick. The operator observes the
effective motion along the surface and modifies the micro-manipulator’s trajectory to
remove position errors. The micro-manipulator controller rejects trajectory modifications
during periods when the operator is unable to view the proper nozzle position.

Under the vertical compensation trajectory, APACTS paints in vertical swaths, where a
swath is a set of horizontal stripes painted sequentially from top to bottom. A stripe
width is the difference between the width of the paint spray and the overlap between
stripes. The experiment's test surface is marked at 30 cm ( 12 in) intervals with horizontal
lines that represent the lower edges of the paint stripes. The micro-manipulator carries a
small camera , whose video signal is displayed on a monitor at the control station. A small

One Degree Micro-Macro Manipulator Integration Test

12

August 2 1 . 2000

arrow is attached to the monitor’s screen to give the operator a reference to assess the
position of the camera relative to the surface (Figure 7). During operation, the arrow
would be positioned at the bottom of the previous stripe to provide the proper overlap.

The operator commands a micro-manipulator trajectory shift through a joystick
potentiometer connected to the micro-manipulator's controller. The joystick input is
scaled to correct the position at up to 7.5 cm/s (3 in/s). Since there is no reference for the
visual feedback during the relocation stroke (i.e., because the lines move relative to the
arrow), the operator's input modifies the trajectory only during the upstroke portion of
the trajectory.

Figure 7. Monitor Screen

3.4 Micro-Macro Interface

The experiments verify a minimally sufficient interface between the micro and macro
manipulators. The interface is the information passed between the controllers of the two
manipulators and the semantics of that data. The information includes the data exchanged
at run-time and any implied data that is encapsulated in the computer code of the
transmitting and receiving controllers. The run-time data consists of the offset to the
micro-manipulator trajectory generated by the operator interface. The implied data
includes the surface direction (always down in the experiment), the data format and the
data frequency.

The micro-macro interface connection is a port on a Seriplex bit bus (from Square D, Inc.)
that serves as the macro-manipulator’s communication and control bus. To fit on the bit
bus, the offset is shifted and scaled to an 8 bit value. The scale and offset values, along
with the implied units, form the interface’s data format and are embedded in software on
both sides of the interface.

One Degree Micro-Macro Manipulator Integration Test

13

August 21, 2000

The data's frequency and regularity follow the requirements and limitations of the micro
and macro manipulators. Data passed across the interface at frequencies greater than the
motion control frequency is unusable by the receiving controller. Generally, a rough order
of magnitude separates the interface frequency (which transfers processed data) and the
motion control frequency (which issues commands). Since the macro-manipulator’s
hydraulic valves limit the macro-manipulator position command frequency to 8 Hz, the
micro-macro interface transfers data at 1.1 Hz in the experiemnts. However, the micro-
manipulator is not able to send data every 0.9 seconds. Since the operator doesn’t have a
visual reference of the proper position relative to the surface during the relocation stroke,
the micro-manipulator does not accept offset changes during the relocation stroke and
there are no reports during those 2.25 seconds. Thus the data pattern consists of five
evenly spaced reports within 3.75 seconds followed by no data for 2.25 seconds. The
interpretation of the offset transmission pattern is embedded in the macro-manipulator
software.

The macro-manipulator controller modifies the boom speed based on the offset data. The
controller scales the current offset, and adds a scaled running sum of previous offsets to
produce the new surface following speed. The controller updates the running sum only
once per micro-manipulator loop to ensure a consistent interval. The macro-manipulator's
acceleration limits smoothly apply the newly computed boom speed. These limits were
not varied during the experiment.

The micro-macro interface transfers the micro-manipulator trajectory offset and trajectory
state. The macro-manipulator controller adjusts the boom speed based on the data and
encoded knowledge of the data format and the data frequency and regularity.

4 Results and Discussions

The experiments test the macro-micro interface by observing the system’s response to
controlled errors. While the system’s performance also depends on the operator's and the
macro-manipulator’s responsiveness, the system response did not need to be optimal to
test the sufficiency of the micro-macro interface.

The system performance depends greatly on the responsiveness of the operator. A rapid
micro-manipulator shift quickly eliminates position errors, produces proper offset values,
appropriately adjusts the macro-manipulator speed, and limits the accumulated errors.
However, an excessively fast interface can cause the operator to overshoot the proper
position and generate erroneous macro-manipulator trajectory changes. The appropriate
values for slow, fast, and too fast depend on the personal preferences of the operator.
After several trials, the test operator ran the experiments with a maximum correction rate
of approximately 7 cm/s (3 in/s).

One Degree Micro-Macro Manipulator Integration Test

14

August 21, 2000

The macro-manipulator speed correction follows the operator's maximum correction rate.
The macro-manipulator speed correction is 0.5 cm/s/cm for the offset and 0.2 (cm/s)/cm
for the running sum of the offset. With the 1. 1 Hz interface frequency, the macro-
manipulator adjusts the boom speed at up to 80% of the operator's maximum correction
rate. Optimal values for the macro-manipulator were not investigated. However, when the
macro-manipulator attempted adjustments at over 100% of the operator’s corrections the
system became unstable.

All experiments were run by the same operator and with the same macro-manipulator
adjustment values. Figure 8 through Figure 1 1 display the results of representative test
runs. These figures show the micro-trajectory offset sent across the interface and the
commanded boom speed. The actual boom speed is subject to the macro-manipulator
acceleration limits and closely follows the commanded speed.

The target lines on the vertical surface represent the lower end of the paint stripes in an
APACTS application. Like the paint stripes the lines are not perfect. Even without
intentional errors, the operator makes numerous corrections to keep the camera on target.
In Figure 8 many small offsets are closely followed by similar adjustments to the boom
speed.

Figure 8. Sample Test Run

5.5
5.4
5.3
5.2
5.1
j 5.0
4.9
4.8
4.7

Time (s)

The first error test began the swath with a 2.5 cm/s ( 1 in/s) boom speed error. The boom
speed error quickly generates a position error that is corrected by the operator through
offsets to the micro-manipulator's trajectory (Figure 9). In response to the operator's
offsets, the macro-manipulator controller decreases the boom speed. The boom speed
reaches its nominal value (5 cm/s, 2 in/s) within approximately one and one-half micro-

One Degree Micro-Macro Manipulator Integration Test

15

August 21, 2000

manipulator loops. The resulting offset (=11 cm) causes the boom to continue to slow to
less than the nominal value. The slower boom speed then generates an opposite position
error that the operator corrects by reducing the micro-manipulator trajectory offset. The
boom speed and the offset slowly return to their nominal values.

Figure 9. Correction with Initial Speed Error

©

©

&

in

E

©

©

cc

Time (s)

The second error test covers a misalignment between the camera and target at the
beginning of the swath. Figure 10 shows the response to an initial 18 cm (7 in) position
error. The operator closes the position error with a similar sized offset within a single
micro-manipulator loop. The system then follows the same pattern as shown in Figure 9
except instead of building a large offset, the existing offset is removed.

Figure 10. Correction with Initial Position Error

One Degree Micro-Macro Manipulator Integration Test

16

August 21 , 2000

A simulated surface experiment shows the control system ability to handle abrupt boom
speed errors (Figure 1 1 ). An interesting aspect of the micro-macro data interface sequence
is seen by comparing the boom speed change at 12 seconds with the one near 72 seconds.
The change at 12 seconds occurred at the beginning of the micro-manipulator upstroke
and was observed and quickly corrected by the operator. The change near 72 seconds
occurred later in the upstroke. The acceleration limit shifts the speed change into the
relocation stroke where the operator can not observe the change. The erroneous speed
was in effect throughout the relocation stroke and caused an error that is still evident in
the micro manipulator offset at the end of the swath (at 90 seconds).

Figure 11. Simulated Surface Following

9.0

8.0

7.0

6.0

5.0

4.0

3.0

2.0

/- s
1/1

E

o

&

E

e

©

A large offset does not equate to a large position error. However a rapidly changing offset
indicates an error correction. The most significant error occurs when a boom speed error
existed during the relocation stroke. Since there isn’t a surface reference during relocation,
the operator can not detect and correct the error. Figure 12 shows the operator’s view
during a relocation sequence when the required boom speed changes by 2.5 cm/s ( 1 in/s) at
the beginning of the relocation stroke. Until near the end of the 2.25 second relocation, the
operator is unaware of the accumulating error. The final 5 cm (2 in) error takes
approximately 0.6 second to correct. The correction could occur in an unpainted
horizontal section or could result in up to 20 cm (8 in) of poorly overlapped paint.

One Degree Micro-Macro Manipulator Integration Test

17

August 21, 2000

Figure 12. Relocation Sequence with Excess Boom Speed

03.75 s 04.42 s 05.28 s 06.00 s

5 Conclusions

A small, quick-response micro-manipulator mounted on a large, sluggish macro-
manipulator produces a manipulator with large volume and good accuracy. The individual
manipulators execute separate trajectories that, when combined, produce a desired tool
path on a surface. Induced errors in the large manipulator’s trajectory test the system’s
response to perturbations and errors. The test results demonstrate a very low bandwidth
interface can coordinate a low bandwidth (i.e., servo frequency) macro-manipulator with
the a high bandwidth (i.e., high servo frequency) micro-manipulator. An interface,
consisting of the micro-manipulator's trajectory offset, the time of the offset, and the
direction of the offset, along with the data format and transmission pattern, is adequate to
maintain control of the system. However, the inability to make corrections during the
relocation stroke of the micro-manipulator's trajectory creates errors that may cause
coverage gaps that would be unacceptable for the APACTS application.

6 Recommendations

The following areas of interface refinement should be investigated. They offer excellent
possibilities of improving the APACTS system.

6.1 Continuous Feedback

The experiment's setup reported offset positions only on the upstroke of the micro-
manipulator trajectory. This creates an irregular interval on which to correct the boom
speed. More responsive action will likely be possible with regular observations of the
trajectory offset. Regular observation would also allow the interface to be more
independent of the macro-manipulator’s program code.

Regular observations can be accomplished by superimposing a target position on the task
position feedback. The observation reference (the alignment arrow in Figure 7) would
shift to the next reference mark (e.g. the bottom of the recent paint stripe) on the monitor
screen at the stall of the relocation stroke, then move on the screen continuously

One Degree Micro-Macro Manipulator Integration Test

18

August 21, 2000

indicating the proper position of the reference mark during the relocation. The operator
(or other sensor) will always have a reference, will always be able to judge the relative
position of the micro-manipulator, and will be able to make corrections throughout the
loop. More frequent corrections should result in smaller errors and faster corrections.

6.2 Surface Modeling

The tests presented in this report presumed the controllers had no information about the
shape of the surface. Information on the curvature could prompt the controller to make
adjustments before an error becomes apparent. Even when the observations and
corrections are not perfect, the remaining error that must be corrected by the operator
would likely be reduced.

7 References

[1] Carderock Division. Naval Surface Warfare Center, SOL N00167-97-SS-R1, “Mechanical Ship Hull
Paint Application System For Use in Drydock”, Commerce Business Daily, April 10, 1997.

[2] T. Yoshikawa, K. Hosoda, T. Doi, H. Murakami, "Dynamic Trajectory Tracking Control of
Flexible Manipulator by Macro-Micro Manipulator System", Proc of ICR A. pp. 1804-1809, 1994.

[3] T. Yoshikawa, K. Harada, A. Matsumoto, "Hybrid Position/Force Control of Flexible-Macro/Rigid-
Micro Manipulator System", IEEE Transactions on Robotics and Automation, Vol. 12, No. 4, Aug
1996.

[4] O. Khatib, "Reduced Effective Inertia in Macro/Mini Manipulator Systems", Proceedings of ACC,
pp. 2140-2147 (1988).

[5] K. Nagai, T. Yoshikawa, "Impedance Control of Redundant Macro-Micro Manipulators", Proc. of
Int'l Conf on Intelligent Robots and Systems, pp. 1438-1445, 1994.

[6] A. Sharon, N. Hogan, D. Hawitt, "High Bandwidth Force Regulation and Inertia Reduction Using a
Macro/Micro Manipulator System", IEEE ICRA, PP. 126-132, 1988.

[7] K. Nagai, Y. Nakagawa, S. IWASA, K. Ohno, "Development of a Redundant Macro-Micro
Manipulator and Contour Tasks Utilizing its Compliant Motion", Proc. of Int'l Conf on Intelligent
Robots and Systems, vol. 1. pp. 279-284, 1997.

[8] T. Narikiyo, H. Nakane, T. Akuta, N. Mohri. N. Saito, "Control System Design, for Macro/Micro
Manipulator with Application to Electrodischarge Machining", Intelligent Robots and Systems, pp
1454-1460, 1994.

[9] J. Albus, H. McCain, R. Lumia , "NASA/NBS Standard Reference Model for Telerobot Control
System Architecture (NASREM)", NIST Technical Report 1235, Gaithersburg, MD 1989.

[10] R. Norcross , "Trajectory Considerations for the Automated Paint Application, Containment, and
Treatment System (APACTS)", NIST Technical Report 6326, Gaithersburg, MD 1999.

One Degree Micro-Macro Manipulator Integration Test

19

August 21, 2000